To dive right into our new version, click here:
You will discover how to:
- Measure AI visibility with consolidated data
- Analyze the sub-queries and sources used by AIs
- Transform AI visibility gaps into marketing actions
- Measure AI reputation across social media and communities
When a customer asks ChatGPT, Gemini or another AI assistant to compare offers, choose a product or understand a market, a brand may be cited, recommended, compared or absent. Manually testing a few questions provides a snapshot, but not a usable measurement: responses can vary from one run to the next.
GEO (Generative Engine Optimization) is the practice of understanding and improving a brand’s visibility in AI-generated answers. LLM Monitor measures that visibility, identifies the sources that contribute to answers and helps prioritize what to do next. Here is what changes in this version.
Measure AI visibility with consolidated data
Language model outputs vary. For the same question, the order of cited brands may change, a brand may appear in one run and disappear in the next, and the wording may differ. Any measurement of AI visibility needs to account for this variability.
To track AI visibility, a single run is not enough to draw conclusions. It captures one answer at one point in time, not how often a brand usually appears. A decision based on a single run can therefore confuse normal model variability with a genuine change in visibility.
LLM Monitor runs each query several times and consolidates the results. This method provides:
- A more robust visibility level, calculated from multiple runs rather than a single answer.
- A clearer view of changes over time: one-off model variations carry less weight when interpreting the indicators.
- A measure of presence frequency: a brand cited once in ten runs does not have the same visibility as a brand cited nine times out of ten.
- Consistent competitor comparisons, because the brand and its competitors are measured using the same method.
We defined this method after observing response variability model by model. The goal is simple: produce a metric that can be tracked over time, not a one-off snapshot of an AI answer.
Analyze the subqueries and sources used by AI systems
When an AI assistant relies on web search, it may launch several intermediate searches based on the initial question. This process of breaking a question into subqueries, known as query fan-out, can cover topics such as price, alternatives, reviews or reliability. The retrieved results are then used to build the final answer.
Showing only the cited sources does not explain how the answer was built. A domain may surface for one specific sub-question without appearing in the same way for others. To take action, each source needs to be connected to the intermediate search that surfaced it.
LLM Monitor now shows, for every question and every run, the observed subqueries, the sources surfaced for each one, the sources actually used to build the answer and the content that belongs to your own domain. This lets you:
- Identify the sub-questions where your brand is absent, especially comparisons, alternatives and reviews.
- Find the domains that help your competitors appear in AI answers while your brand is missing.
- Distinguish your content that was merely surfaced from content that was actually used to build the answer.
The diagnosis becomes more precise: instead of simply knowing that your AI visibility is low, you can see which subqueries your brand is missing from and which sources contribute to your competitors’ presence.
Turn AI visibility gaps into marketing actions
Measuring AI visibility only matters if the diagnosis leads to action. After an audit, the useful question is simple: which actions can the marketing team actually implement?
LLM Monitor follows one simple rule: an opportunity must be actionable by a marketing or communications team, without requiring heavy technical work. Recommendations focus on creating or improving content, both on your own channels and on third-party platforms: review sites, partner blogs, forums, video content, social networks and specialist publications.
Each opportunity includes the information needed to prioritize it:
| Information available | How the team can use it |
|---|---|
| Data-based rationale | Connect the action to a query, a source or a visibility gap with a competitor |
| Difficulty and estimated effort | Prioritize quick wins and larger projects |
| Opportunity status | Track an action as to do, implemented or ignored, with its implementation date |
| History across audits | Keep previously suggested actions so the plan can evolve without starting from scratch |
| Before-and-after trend | Compare indicators before and after an action is implemented |
The methodology needs to remain rigorous here: LLM Monitor measures an observed correlation, not proven causation. An increase in visibility after an action does not prove that the action was the sole cause. The platform provides a before state, an after state and consolidated data so changes can be compared without overinterpreting the result.
Measure AI reputation across social media and communities
AI visibility measures whether a brand is present in answers. AI reputation measures how that brand is described in those answers. LLM Monitor already covered press coverage, consumer reviews and an overall view. It now includes a dedicated scope for social networks, forums, communities and video content.
The new Social & Communities dimension has its own score, summary, and identified strengths and weaknesses. It is included in the composite AI reputation index. At the same time, the consumer dimension is now limited to review platforms so that each source category has a clearly defined scope.
This separation helps identify where a reputation signal comes from. A negative perception found in forums does not call for the same actions as a negative perception coming from the press or review platforms.
Query LLM Monitor data from an AI assistant
LLM Monitor can now be connected to an AI assistant so you can query your account data in natural language. This means you can access an analysis without opening the dashboard for every question.
For example, you can ask: “Summarize the AI reputation from my latest analysis,” “Is my analysis complete, and what are the main results?” or “List the sites that cite my two main competitors but not my brand, then suggest an article topic.” The goal is to use LLM Monitor data directly from the tool where you already work.
A clearer interface and AI models tracked over time
The dashboard has been redesigned to make indicators easier to read and compare. The new interface standardizes KPIs, aligns figures across views, adds new metrics and introduces an audit page where you can find historical results by project, brand and month.
Model coverage is also updated as new versions become available, including free consumer versions. If a model used in a previous report is no longer available, LLM Monitor flags it before a new analysis is launched so comparisons over time remain easy to interpret.
What this version changes for AI visibility management
Counting mentions is not enough to manage a brand’s visibility in AI answers. You need to measure how often the brand appears, understand which subqueries and sources contribute to the answers, compare the brand with its competitors and turn the observed gaps into actions that can be tracked over time.
That is the role of LLM Monitor: connecting measurement, diagnosis and action in one tool.
You can test LLM Monitor directly on your own market. In about 20 minutes, you get an initial view of your brand’s visibility versus its competitors, the associated sources and the first GEO opportunities.
Frequently asked questions about AI visibility and LLM Monitor
How can you reliably measure a brand’s visibility in AI answers?
Language model outputs vary from one run to another. A single answer can therefore overestimate or underestimate a brand’s presence. LLM Monitor repeats queries and consolidates the results to measure presence frequency on a more robust basis.
What is query fan-out?
Query fan-out is the process of launching several intermediate searches from an initial question. Seeing these subqueries and their sources helps identify the topics where your brand is absent and the ones that make your competitors appear.
How does LLM Monitor turn an AI visibility audit into actions?
LLM Monitor turns observed visibility gaps into actionable opportunities. Each one is designed to be implemented by a marketing or communications team, with a difficulty level and tracking of its implementation.
Can an increase in AI visibility be attributed to a specific GEO action?
You can observe correlation, not prove causation. LLM Monitor compares indicators before and after an action using consolidated data, without automatically attributing the change to a single factor.
Questions related to this article
How can you reliably measure a brand's visibility in AI answers?
Language model outputs vary from one run to another. A single answer can therefore overestimate or underestimate a brand's presence. LLM Monitor repeats queries and consolidates the results to measure presence frequency on a more robust basis that can be compared over time.
What is query fan-out?
Query fan-out is the process of launching several intermediate searches from an initial question. Seeing these subqueries and the sources surfaced for each one helps explain which topics and domains contribute to the final answer, and where the brand is absent compared with its competitors.
How does LLM Monitor turn an AI visibility audit into actions?
LLM Monitor turns gaps observed across queries, sources and competitor visibility into actionable opportunities. Each opportunity is designed to be implemented by a marketing or communications team, with a difficulty level and tracking of its implementation.
Can an increase in AI visibility be attributed to a specific GEO action?
Not with certainty. LLM Monitor can show a correlation between an action and a change in indicators, but it cannot prove that a single factor caused the change. Before-and-after comparisons help measure the change without overinterpreting its cause.
